Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via...
Transcript of Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via...
![Page 1: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/1.jpg)
Modeling and characterizing anisotropic inclusion orientation in heterogeneousmaterial via directional cluster functions and stochastic microstructure reconstructionYang Jiao and Nikhilesh Chawla Citation: Journal of Applied Physics 115, 093511 (2014); doi: 10.1063/1.4867611 View online: http://dx.doi.org/10.1063/1.4867611 View Table of Contents: http://scitation.aip.org/content/aip/journal/jap/115/9?ver=pdfcov Published by the AIP Publishing Articles you may be interested in Characterizing randomly anisotropic surfaces in eddy-current NDE AIP Conf. Proc. 1430, 340 (2012); 10.1063/1.4716248 RECENT ADVANCES IN MODELING DISCONTINUITIES IN ANISOTROPIC AND HETEROGENEOUSMATERIALS IN EDDY CURRENT NDE AIP Conf. Proc. 1335, 1565 (2011); 10.1063/1.3592116 Surface and microstructural characterization of laser beam welds in an aluminum alloy J. Vac. Sci. Technol. A 20, 1416 (2002); 10.1116/1.1487868 Role of anisotropy in noncontacting thermoelectric materials characterization J. Appl. Phys. 91, 225 (2002); 10.1063/1.1416852 Anisotropic effective conductivity of materials with nonrandomly oriented inclusions of diverse ellipsoidal shapes J. Appl. Phys. 87, 8561 (2000); 10.1063/1.373579
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 2: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/2.jpg)
Modeling and characterizing anisotropic inclusion orientation inheterogeneous material via directional cluster functions and stochasticmicrostructure reconstruction
Yang Jiaoa) and Nikhilesh ChawlaMaterials Science and Engineering, Arizona State University, Arizona 85287-6206, USA
(Received 31 October 2013; accepted 22 February 2014; published online 6 March 2014)
We present a framework to model and characterize the microstructure of heterogeneous materials
with anisotropic inclusions of secondary phases based on the directional correlation functions of
the inclusions. Specifically, we have devised an efficient method to incorporate both directional
two-point correlation functions S2 and directional two-point cluster functions C2 that contain
non-trivial topological connectedness information into the simulated annealing microstructure
reconstruction procedure. Our framework is applied to model an anisotropic aluminum alloy and
the accuracy of the reconstructed structural models is assessed by quantitative comparison with the
actual microstructure obtained via x-ray tomography. We show that incorporation of directional
clustering information via C2 significantly improves the accuracy of the reconstruction. In addition,
a set of analytical “basis” correlation functions are introduced to approximate the actual S2 and C2
of the material. With the proper choice of basis functions, the anisotropic microstructure can be
represented by a handful of parameters including the effective linear sizes of the iron-rich and
silicon-rich inclusions along three orthogonal directions. This provides a general and efficient
means for heterogeneous material modeling that enables one to significantly reduce the data set
required to characterize the anisotropic microstructure. VC 2014 AIP Publishing LLC.
[http://dx.doi.org/10.1063/1.4867611]
I. INTRODUCTION
An accurate knowledge of the microstructure of a
heterogeneous material is crucial for subsequent modeling
and prediction of its physical properties and performance.
In the past few decades, a quantitative understanding of
the structure-property relation in such heterogeneous materi-
als has begun to emerge, mainly due to the development of
advanced experimental and computational material micro-
structure characterization techniques.1–3 In particular, x-ray
tomography technique,4–6 which eliminates destructive
cross-sectioning and allows for superior resolution and
image quality with minimal sample preparation,7,8 has been
widely used to obtain high-resolution three-dimensional
(3D) microstructure for a wide range of heterogeneous mate-
rials, including Sn-rich alloys,9 powder metallurgy steels,10
metal matrix composites,11–15 and lightweight alloys.16–20
The resulting microstructural data sets can be subsequently
incorporated into finite element models to predict the onset
of local damage mechanisms and the macroscopic deforma-
tion behavior.21–25
Experimentally obtained microstructural data are
usually represented as a large array whose entries indicate
the local states of the microstructure (e.g., the phase that a
specific voxel belongs to). Although morphological details
of a specific material can be contained in its microstructural
array, such information would not all be useful for macro-
scopic effective property analysis, which requires averaging
over a sufficiently large number of material samples to yield
meaningful and robust statistics. The structural details asso-
ciated with a specific material sample usually do not signifi-
cantly contribute to the overall statistics. In addition, in the
case where the entire course of microstructural evolution is
of interest (e.g., during the solidification or coarsening
processes), the resulting 4D structural data sets (three spatial
dimensions plus one temporal dimension) are usually
extremely large.
Therefore, it is highly desirable to devise a robust frame-
work that enables one to represent, characterize, and model
the complex microstructure of a heterogeneous material in a
much more efficient manner. Recently, it has been suggested
that certain statistical morphological descriptors, such as
lower-order spatial correlation functions of the material’s
phases, can be employed to model a wide spectrum of com-
plex microstructures.26–30 This allows one to reduce the large
data sets for a complete specification of all of the local states
in a microstructure to a handful of simple scalar functions
that statistically capture the salient structural features.
Stochastic reconstruction techniques such as the simulated
annealing procedure developed by Yeong and Torquato31,32
can then be employed to generate realistic virtual 3D micro-
structures from the correlation functions. Material modeling
and reconstruction based on correlation functions has
enabled the development of data-driven material design
schemes33 and structure-based performance modeling.34
In this paper, we present a framework to model hetero-
geneous materials with anisotropic inclusions of secondary
phases based on the directional correlation functions of the
inclusions. In particular, we devise an efficient method that
a)Author to whom correspondence should be addressed. Electronic mail:
0021-8979/2014/115(9)/093511/9/$30.00 VC 2014 AIP Publishing LLC115, 093511-1
JOURNAL OF APPLIED PHYSICS 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 3: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/3.jpg)
enables us to incorporate directional two-point cluster func-
tions C2 into the reconstruction procedure, which contain
non-trivial topological connectedness information of the
material’s phases.28,35 We apply our general method to
model the microstructure of an Al alloy with anisotropic
inclusions of Fe-rich and Si-rich phases. We show quantita-
tively that incorporating such clustering information can lead
to an improved reconstruction by comparing the lineal-path
functions of the actual and reconstructed microstructures. In
addition, we introduce a set of analytical “basis” correlation
functions to approximate the actual S2 and C2 of the material.
This enables us to further reduce the set of parameters
required to characterize the anisotropic material microstruc-
ture. Specifically, we show that with our choice of basis
functions, a heterogeneous microstructure can be represented
by a handful of parameters including the effective linear
sizes of the Fe-rich and Si-rich inclusions along three orthog-
onal directions.
The rest of the paper is organized as follows: In
Sec. II, we provide definitions of the statistical morpho-
logical descriptors (i.e., correlation functions) employed
to model the alloy microstructure. In Sec. III, we present
the Yeong-Torquato (YT) reconstruction procedure and
our method for efficient sampling of directional correla-
tion functions from trial microstructures. In Sec. IV, we
employ the functions S2 and C2 and the YT reconstruction
procedure to model an anisotropic alloy microstructure.
In addition, we introduce a set of analytical “basis” corre-
lation functions associated with elongated particles to
approximate the actual S2 and C2 of the material to fur-
ther reduce the set of parameters required to characterize
the heterogeneous microstructure. Concluding remarks
are given in Sec. V.
II. STATISTICAL MICROSTRUCTURAL DESCRIPTORS
A. Two-point correlation function
In general, the microstructure of an Al alloy with aniso-
tropic secondary phase inclusions in the Al matrix can be
uniquely determined by specifying the indicator functions
associated with all of the individual phases (the inclusions
and the matrix),1 i.e.,
I ið ÞðxÞ ¼1 x in phase i
0 otherwise;
((1)
where i¼ 1,…, q and q is the total number of phases. The
volume fraction of phase i is then given by
ui ¼ hI ið ÞðxÞi; (2)
where hi denotes the ensemble average over many independ-
ent material samples or volume average over a single large
sample if it is spatially “ergodic.”1 The two-point correlation
function SðijÞ2 ðx1; x2Þ associated with phases i and j is defined
as
SðijÞ2 ðx1; x2Þ ¼ hI ið Þðx1ÞI jð Þðx2Þi (3)
which also gives the probability that two randomly selected
points x1 and x2 fall into phases i and j, respectively
(Fig. 1(a)). For a material with q distinct phases, there are
totally q� q different S2. However, it has been shown that
only q of them are independent1,36 and the remaining q� (q� 1) functions can be explicitly expressed in terms of
the q independent ones. In our case, q¼ 3 (Fe-rich inclu-
sions, Si-rich inclusions, and the Al matrix) and thus, we
only need to consider the two-point correlation functions
associated with the anisotropic inclusions, including the
auto-correlation function of the Fe-rich inclusions SFe2 , the
auto-correlation function of the Si-rich inclusions SSi2 , and
cross correlation function SFe�Si2 .
For statistically homogeneous materials such as the Al
alloy considered here, there is no preferred center in the
microstructure. Therefore, the associated S2 depends only on
the relative vector displacement between the two points,1
i.e.,
S2ðx1; x2Þ ¼ S2ðx2 � x1Þ ¼ S2ðrÞ; (4)
where r¼ x2� x1. At r¼ 0, the auto-correlation function
gives the probability that a randomly selected point falls into
the phase of interest, i.e., the volume fraction of the associ-
ated phase. However, the cross correlation function is zero at
r¼ 0, since the probability of finding a single point falling
into two different phases is zero. At large r values, the
FIG. 1. Schematic illustration of the events that contribute to various correlation functions. The two-point correlation function S2 gives the probably of finding
two points in the phases of interest. In (A), we show events that contribute to S2(ij), where i, j can be either red phase or blue phase. The lineal-path function
L(r) gives the probability that a randomly chosen line segment of length r entirely falls into the phase of interest. In (B), we show events that contribute to L(i).
The two-point cluster function C2(r) gives the probably of finding two points separated by r in the same cluster of the phase of interest. In (C), we show events
that contribute to C2(i).
093511-2 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 4: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/4.jpg)
probabilities of finding the two points in the phases of inter-
est are independent of one another, leading to u2i for the
auto-correlation functions and uiuj for the cross correlation
function.
For a statistically isotropic material, S2 depends only on
the scalar distance between a pair of points. In the case of
the Al alloy, the secondary phase inclusions are significantly
elongated along the rolling direction (RD), leading to an
overall anisotropic microstructure with a preferred direction.
Therefore, we employ the S2 along three orthogonal direc-
tions, i.e., the RD, transverse (TD), and normal (ND) direc-
tions, to characterize the microstructure. Accordingly, we
denote the directional two-point correlation function of type
a along direction b by Sa;b2 ðrÞ, where a¼ {Fe, Si, Fe-Si} and
b¼ {RD, TD, ND}. We note that the general n-point correla-
tion function Sn which gives the probability of finding a
particular n-point configuration in specific phases can be
defined in a similar manner as S2; see Eq. (2). It has been
shown that the effective properties of a heterogeneous mate-
rial can be explicitly expressed as series expansions involv-
ing certain integrals of Sn. Interested readers are referred to
Ref. 2 for detailed discussions of Sn and their properties.
B. Lineal-path function
The lineal-path function L(i)(r) gives probability that a
randomly selected line segment of length r¼ |r| along the
direction of vector r entirely falls into phase i (Fig. 1(b)).1,37
At r¼ 0, L(i)(0) reduces to the probability of finding a point
in phase i and thus, L(i)(0)¼ui. In materials that do not con-
tain system-spanning clusters, the chance of finding a line
segment with very large length entirely falling into any
phases is vanishingly small. Accordingly, for large r values,
L(i) decays to zero rapidly in such materials. The lineal-path
function contains partial topological connectedness informa-
tion of the material’s phases, i.e., that along a lineal-path.
Generally, the lineal-path function underestimates the degree
of clustering in the system (e.g., two points belonging to the
same cluster but not along a specific lineal path will not con-
tribute to L). For the anisotropic alloy microstructure consid-
ered here, we use the directional lineal-path functions La;bðrÞfor the Fe-rich and Si-rich inclusions along three orthogonal
directions, where a¼ {Fe, Si} and b¼ {RD, TD, ND}.
C. Two-point cluster function
The two-point cluster correlation function CðiÞ2 ðx1; x2Þ
gives the probability that two randomly selected points x1 and
x2 fall into the same cluster of phase i (Fig. 1(c)).1,35 A cluster
of a phase is defined as a compact region in which any point
in the region is connected to any other point in the region by a
continuous path completely in the region. For statistically ho-
mogeneous materials, C2 depends only on the relative vector
displacement between the two points, i.e., C2ðx1; x2Þ ¼ C2ðrÞ.In contrast to the lineal-path function, C2 contains complete
clustering information of the phases, which has been shown to
have dramatic effects on the material’s physical properties.1
Moreover, unlike S2 and L, the cluster functions generally can-
not be obtained from lower-dimensional cuts (e.g., 2D slices)
of a 3D microstructure, which may not contain correct con-
nectedness information of the actual 3D system.
It has been shown that C2 is related to S2 via the follow-
ing equation:35
SðiiÞ2 ðrÞ ¼ C
ðiÞ2 ðrÞ þ D
ðiÞ2 ðrÞ; (5)
where DðiÞ2 ðrÞ measures the probability that two points sepa-
rated by r fall into different clusters of the phase of interest.
In other words, C2 is the connectedness contribution to the
standard two-point correlation function S2. As in the case of
S2 and L, we employ directional cluster functions Ca;b2 ðrÞ to
characterize the anisotropic alloy, where a¼ {Fe, Si} and
b¼ {RD, TD, ND}. For microstructures with well-defined
inclusion, C2(r) of the inclusions is a short-ranged function
that rapidly decays to zero as r approaches the largest linear
size of the inclusions. We note that although C2 is a
“two-point” quantity, it has been shown to embody
higher-order structural information which makes it a highly
sensitive statistical descriptor over and above S2.28,38
III. STOCHASTIC MICROSTRUCTURERECONSTRUCTION PROCEDURE
A. Simulated annealing procedure
We use the YT reconstruction procedure31,32 to generate
virtual 3D microstructures from a specific set of directional
correlation functions discussed in the previous section. We
note that there are many other different microstructure recon-
struction procedures, such as the Gaussian random field
method,39 phase recovery method,40 and the recently devel-
oped raster path method.41 However, the YT procedure ena-
bles one to incorporate an arbitrary number of correlations of
any type, while the others require specific structural informa-
tion as input.
In the YT procedure, the reconstruction problem is
formulated as an “energy” minimization problem, with the
energy functional E defined as follows:
E ¼X
a
Xb
Xr
f a;bðrÞ � �fa;bðrÞ
h i2
; (6)
where �fa;bðrÞ is a target correlation function of type a along
direction b and f a;bðrÞ is the corresponding function associ-
ated with a trial microstructure. The simulated annealing
method42 is usually employed to solve the aforementioned
minimization problem. Specifically, starting from an initial
trial microstructure (i.e., old microstructure) which contains
a fixed number of voxels for each phase consistent with the
volume fraction of that phase, two randomly selected voxels
associated with different phases are exchanged to generate a
new trial microstructure. Relevant correlation functions are
sampled from the new trial microstructure and the associated
energy is evaluated, which determines whether the new trial
microstructure should be accepted or not via the probability
paccðold!newÞ¼min 1; expEold
T
� ��exp
Enew
T
� �( ); (7)
093511-3 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 5: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/5.jpg)
where T is a virtual temperature that is chosen to be high ini-
tially and slowly decreases according to a cooling sched-
ule.31,32 The above process is repeated until E is smaller than
a prescribed tolerance, which we choose to be 10�10 here.
Generally, several hundred thousand trials need to be made
to achieve such a small tolerance.
B. Efficient sampling of correlation functions fromtrial microstructure
It can be clearly seen that the key step in the reconstruc-
tion process is to efficiently compute the correlation
functions from the trial microstructure. For the alloy micro-
structure, we incorporate the directional two-point correla-
tion function S2 and two-point cluster function C2 into the
reconstructions. The directional lineal-path function L is
employed to ascertain the accuracy of the reconstructions.
Since an extremely large number of trial microstructures are
generated, it is highly undesirable to completely re-compute
the functions for each trial. Instead, one can use the correla-
tion functions of the old microstructure to obtain those for
the new microstructure, which we discuss in detail in this
section.
In the reconstructions, we only consider S2 and C2 along
three orthogonal directions. Accordingly, the original 3D
microstructural array can be decomposed into three subsets
of 1D arrays. The 1D arrays in each subset are along one of
the three orthogonal directions. The correlation functions of
interest associated with the 1D arrays in a subset are com-
puted and averaged to obtain the corresponding directional
correlation function. Specifically, we consider the voxels in a
1D array as “particles” of a lattice-gas system. For the auto-
correlation functions, the voxels of the phase of interest are
grouped and the separation distance between a pair of voxels
within the group is computed. These distances are then
binned and normalized by the size of the 1D array to obtain
the corresponding correlation function. To obtain the
cross-correlation functions, the voxels of two different
phases are grouped and the separation distance between a
pair of voxels, one from each group, is computed, binned,
and normalized. For the cluster functions, the voxels belong-
ing to the same cluster of the phase of interest are grouped.
The pair separation distances within each group are com-
puted, binned, and normalized to obtain the directional C2.
In the generation of a new trial microstructure, the posi-
tions of a pair of randomly selected voxels belonging to
different phases are switched. For S2, only six 1D micro-
structural arrays, two from each subset for a specific direc-
tion, are affected. Accordingly, only the directional
correlation functions associated with these six 1D micro-
structural arrays need to be re-computed in order to obtain
the directional two-point correlation functions of the entire
new trial microstructure. This enables one to efficiently
re-compute S2 for the trial microstructures. For C2, one needs
to track the evolution of the clusters due to the voxel switch-
ing. There are three possible “cluster events” associated with
the voxel switching: (i) creation of a new cluster of a phase
(i.e., an isolated voxel surrounded by voxels of other phases);
(ii) combination of small clusters into a single cluster; (iii)
breaking of a cluster into small clusters. After each switch-
ing, possible cluster events are checked and the cluster con-
figuration in the microstructure is updated accordingly. The
directional C2 are then re-computed based on the updated
cluster configuration. Note that in general more than six 1D
arrays will be affected by the cluster events. Therefore,
reconstructions incorporating C2 are more time-consuming
than the S2-alone reconstructions.
IV. RESULTS
In the section, we will present our framework for model-
ing anisotropic heterogeneous materials via the directional
two-point correlation function S2 and directional two-point
cluster function C2, which is motivated by a number of stud-
ies showing that using separate structural information along
three orthogonal directions can lead to accurate reconstruc-
tions of anisotropic microstructures.29,43–46 Specifically, the
utility of the framework will be illustrated by applying it to
model an anisotropic alloy microstructure, which was
obtained via high-resolution x-ray tomographic microscopy
(Fig. 2). Aluminum (Al) alloys have been widely used in
many engineering structures especially in automotive and air-
craft applications due to their unique high strength-to-weight
ratio and corrosion resistance. An aluminum alloy almost
always has secondary phase inclusions and particles with
impurities that are present in the microstructure. Due to roll-
ing of the alloy, the inclusions usually process anisotropic
shapes, with the dimension along the rolling direction signifi-
cantly elongated. This in turn results in an overall anisotropic
alloy microstructure.
By quantitatively comparing various reconstructed
microstructures from different correlation functions with the
actual microstructure, we show that incorporating clustering
information can further improve the accuracy of the
FIG. 2. 3D alloy microstructure with anisotropic inclusions of Fe-rich
(shown in green) and Si-rich (shown in red) phases in Al matrix (transpar-
ent) obtained from x-ray tomographic microscopy. The size of the sample is
roughly 360 lm� 680 lm� 860 lm. This figure is reproduced by permis-
sion from Singh et al., Metall. Mater. Trans. A 43, 4470–4474 (2012).
Copyright 2012 by Springer Science & Business Media.
093511-4 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 6: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/6.jpg)
reconstruction, although S2-alone reconstruction already
capture the salient morphological features of the alloy micro-
structure. In addition, we introduce a set of analytical “basis”
functions associated with elongated particles to approximate
the S2 and C2 of the inclusions. This allows us to characterize
the alloy with only six effective linear sizes of the Fe-rich
and Si-rich inclusions along three orthogonal directions.
A. Microstructure modeling
Figure 2 shows the 3D microstructure of the alloy
obtained from image segmentation and reconstruction of the
x-ray tomography data set. It can be clearly seen that both
the Fe-rich and Si-rich inclusions are elongated along the
RD. On the other hand, the inclusions along the TD and
ND directions are more isotropic. These result in an overall
anisotropic microstructure. Details of the x-ray tomography
process, conducted at the Advanced Photon Source, Argonne
National Laboratory, are reported elsewhere.47
Figure 3 shows the directional two-point correlation func-
tion S2 and two-point cluster function C2 associated with the
inclusion phases. The functions associated with the actual
microstructure are sampled from the complete 3D microstruc-
tural array in order to correctly capture the clustering informa-
tion. It can be seen that the autocorrelation functions SFe2 and
SSi2 almost monotonically decay to the associated long-range
asymptotic values ui2. Moreover, the cross-correlation func-
tion SFe�Si2 rapidly increases to the associated asymptotic
value uiuj. These behaviors suggest that inclusions almost
have no spatial correlations between one another, i.e., they are
nearly randomly distributed in the microstructure. This is
because the volume fractions of the inclusions are extremely
small, and thus, the system can be considered at the low ulimit. As we will quantitatively show below, at this limit, the
microstructure is mainly determined by the distribution of the
shape and size of the inclusions. The two-point cluster func-
tion C2 associated with the inclusion phases monotonically
decay to zero. Similar to the autocorrelation functions, the
range of C2 along the RD is much longer than the other two
directions, clearly indicating that the clusters (inclusions) are
elongated along this direction.
Here, we approximate the autocorrelation function of
the inclusion phase a with the following function:
Sa;b2 ðrÞ ¼ ua Aa
b expðBab � rÞ þ ð1�Aa
bÞVintðr; �DabÞ
h inþu2
a 1�Vintðr; �DabÞ
h ioHð �D
ab � rÞ þu2
aHðr� �DabÞ;
(8)
where ua is the volume fraction, Aab, Ba
b, and �Dab are parame-
ters to be determined. The approximation (9) includes two
“basis” functions: the Debye random medium function
FIG. 3. Directional correlation functions associated with the inclusion phases in the Al alloy. (A) Directional two-point correlation functions SFe2 of the Fe-rich
inclusions. (B) Directional two-point correlation functions SSi2 of the Si-rich inclusions. (C) Directional cross-correlation functions SFe�Si
2 associated with the
inclusion phases. (D) Directional two-point cluster functions CFe2 of the Fe-rich inclusions. (E) Directional two-point cluster function CSi
2 of the Si-rich inclu-
sions. The unit of length in the plots is the linear size of a cubic voxel, which is �1.8 lm.
093511-5 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 7: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/7.jpg)
exp(Br)26,48 and Vint(r; D),1 which is the scaled intersection
volume of two isotropic inclusions with linear size D sepa-
rated by r:
Vintðr; DÞ ¼ 1� 3r
2Dþ r3
2D3(9)
and HðxÞ is the Heaviside function, i.e.,
HðxÞ ¼1 x � 0
0 x < 0:
((10)
In other words, one can consider the alloy microstructure as
a mixture of “clusters of all shape sizes” (features of a
Debye random medium) and well defined inclusions (charac-
terized by Vint). For the anisotropic inclusions, we consider
that each direction possesses distinct combination parameter
Aab, correlation parameter Ba
b, and effective size �Dab, which is
the effective linear size or “length” of the inclusions along
that direction. The values of these parameters are determined
such that the approximated functions best match the actual
autocorrelation function (i.e., the squared difference between
two functions is minimized) and the value of the effective
inclusion lengths are obtained via this fitting procedure. In
Tables I and II, we provide the values of the parameters for
the Fe-rich and Si-rich inclusions in the alloy, respectively.
Note that we do not model the cross-correlation function
SFe�Si2 since at the low u limit its values are several orders of
magnitude smaller than the autocorrelation functions, which
again suggests that there are no significant inter-particle spa-
tial correlations.
By definition, the two-point cluster function C2 meas-
ures the contributions from the point pairs in the same cluster
(inclusion) in S2; see Eq. (5). Thus, we approximate C2 of
the inclusion phase a as follows:
Ca;b2 ðrÞ ¼ua Aa
b expðBab � rÞ þ ð1� Aa
bÞ 1� 3r
2 �Dab
þ r3
2 �Dab
!" #
�Hð �Dab � rÞ; ð11Þ
where the parameters Aab, Ba
b, and �Dab are given in Tables I
and II, respectively, for the Fe-rich and Si-rich inclusions.
We note that these parameters are identical for S2 and C2,
because C2 is the connectedness portion of S2. Thus, the
approximated C2 does not provide additional mathematically
information concerning the material microstructure besides
that contained in corresponding S2. However, as we will
show in the next subsection, incorporating C2 in the recon-
struction procedure to constrain clustering in the
microstructure will improve the accuracy of the reconstruc-
tions. In addition, we emphasize that although the S2 and C2
appear to be close to one another in Fig. 3 (due to the small
inclusion volume fractions), these are intrinsically different
quantities in that S2 is long-ranged and C2 decays to zero
beyond the largest linear inclusion size.
We note that in Eqs. (8) and (11), a number of “basis”
functions have been employed to approximate the correlation
functions of the material. It has been shown26,36 that the
collection of all realizable standard two-point correlation
functions S2 span a well-defined correlation function space.
This implies that in principle, the standard two-point correla-
tion function of an arbitrary heterogeneous material can be
exactly represented as the linear combination of a complete
set of basis functions. In practice, such a complete set of
basis functions are extremely difficult to obtain (if not
impossible) and thus, an incomplete subset of basis functions
(including the ones used here) have been suggested,26 which
are associated with well known model microstructures with
distinct salient morphological features. Although empirically
chosen, the subset of basis functions has been shown to be
able to represent a wide spectrum of heterogeneous material
microstructures26,27 and allows one to “decompose” the
structural features of the material of interested into a set of
key features associated with very known model structures.
The space of two-point cluster functions C2 has not been
systematically investigated before. However, as indicated in
Sec. II C, C2 is the connectedness contribution to the standard
two-point correlation function S2. Therefore, it can be
expected that the connectedness portions of the basis functions
for the S2 space provide a natural choice of the basis functions
for the C2 space. Here, we have used the intersection-volume
function Vint(r) as one of the basis functions to approximate C2
of the inclusion phases, which is the connected contribution to
the S2 of a binary material with dilute distribution of
non-overlapping particles in a matrix. A rigorous mathemati-
cal treatment of the correlation function space is out of scope
of the current paper and interested readers are referred to Refs.
26 and 36 and references therein for detailed discussions.
B. Microstructure reconstruction
We employ the stochastic reconstruction procedure
discussed in Sec. III to determine the minimal set of structure
information required to statistically characterize the aniso-
tropic alloy microstructure. Specifically, we consider the
reconstructions from the directional S2 alone and from the
combination of directional S2 and C2, including those
obtained from the experimental data and the analytical
approximation. We note that only autocorrelation functions
TABLE I. The values of Aab, Ba
b, and �Dab for the Fe-rich inclusions along
three orthogonal directions in the alloy.
Fe-rich inclusions RD TD ND
Aab 0.801 0.850 0.902
Bab �0.350 �0.504 �0.579
�Dab 81.2 32.0 18.9
TABLE II. The values of Aab, Ba
b, and �Dab for the Si-rich inclusions along
three orthogonal directions in the alloy.
Si-rich inclusions RD TD ND
Aab 0.918 0.898 0.951
Bab �0.480 �0.548 �0.679
�Dab 86.8 25.9 11.0
093511-6 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 8: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/8.jpg)
of the inclusions are used for the reconstructions. Figure 4
shows the visual comparison of a portion of the actual alloy
microstructure with various reconstructions. The microstruc-
tures are visually indistinguishable from one another.
Figure 5 shows the lineal-path functions L of the Fe-rich
and Si-rich inclusions in the actual and reconstructed micro-
structures. This quantity is employed to ascertain the accuracy
of the reconstructions because it contains lineal connectedness
information of the inclusion phases, and thus, is sensitive the
degree of clustering in the system. It can be seen that L associ-
ated with different microstructures lie closely to one another,
which is consistent with the visual comparison. However,
quantitative differences can still be observed. The lineal-path
functions of the S2-alone reconstruction are slightly longer
ranged than those of the actual microstructure, indicating
slight overestimation of the degree of clustering in the recon-
struction, which is consistent with the observation report in
Ref. 29. On the other hand, reconstructions incorporating C2
can reproduce the degree of clustering very well, as measured
by L. This is not surprising since C2 contains the complete
connectedness information while L only contains partial infor-
mation, as we discussed in Sec. II.
Importantly, one can see that the reconstruction from the
approximated S2 and C2 can very well reproduce the salient
structural features of the alloy. This suggests that the alloy
microstructure can be statistically characterized by the pro-
posed analytical approximations of the correlation functions,
which depend only on parameters given in Tables I and II. In
other words, with our choice of the basis functions, the 3D
microstructural array specifying the local state of each indi-
vidual voxels can be reduced to a handful of scalar parame-
ters that statistically model the alloy microstructure.
V. SUMMARY AND CONCLUDING COMMENTS
We have provided a framework to model the microstruc-
ture of a heterogeneous material with anisotropic inclusions
secondary phases via directional correlation functions of the
inclusions and stochastic reconstructions. Specifically, we
proposed a set of “basis” functions parameterized by the
effective linear size of the inclusions to approximate the
directional two-point correlation function S2 and two-point
cluster function C2 of the inclusion phases. By devising an
efficient method that enables us to incorporate directional C2
into the reconstruction procedure, we quantitatively showed
that the analytical approximations can lead to reconstructions
that very well reproduce the salient structural features of the
material. This suggests that the heterogeneous microstructure
can be statistically modeled by a handful of parameters,
which significantly reduces the set of parameters required to
characterize the material.
In general, directional S2 alone is not sufficient to char-
acterize a microstructure even in a statistical sense.49,50 It
has been well established that S2-alone reconstructions
FIG. 4. Visual comparison of the actual alloy microstructure with various reconstructions. The Fe-rich inclusions are shown in blue and Si-rich inclusions in
red. The Al matrix is transparent. The size of samples is 102� 192� 248 voxels (�180 lm� 340 lm� 430 lm). (A) A portion of the actual alloy microstruc-
ture obtained from tomography. (B) S2 alone reconstruction. (C) S2-C2 reconstruction from the correlation functions associated with the actual microstructure.
(D) S2-C2 reconstruction from the approximated correlation functions.
FIG. 5. Quantitative comparison of the
directional lineal-path functions of the
actual alloy microstructure and various
reconstructions. (A) Fe-rich inclusions.
(B) Si-rich inclusions. The unit of
length in the plots is the linear size of a
cubic voxel, which is �1.8 lm.
093511-7 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 9: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/9.jpg)
usually significantly overestimate the degree of clustering in
the system.30 Thus, the alloy microstructure studied here is a
special case where S2 alone can lead to fairly accurate recon-
structions. This is because the system is in the low
volume-fraction limit. During the reconstruction process, it
is very difficult to form large clusters due to the small num-
ber of voxels associated with the inclusion phases.
Nonetheless, the reconstruction can be further improved by
incorporating the cluster function C2. At the low u limit,
there are not significant spatial correlations between the
inclusions, and thus, the cross-correlation function becomes
irrelevant to the reconstruction. For systems at higher vol-
ume fraction, inter-phase correlations become crucial to
determining the microstructure and physics-based models
(e.g., hard-particle packings51) should be employed to deter-
mine and approximate such correlations.
The general framework to model complex microstruc-
tures via correlation functions can be applied to other hetero-
geneous materials such as composites and porous media as
well. Specifically, the two-point correlation function space
alone has been shown to be useful for the modeling of a vari-
ety of complex microstructures.26,27 It can be expected that
the incorporation of additional correlation functions can lead
to more accurate characterization and modeling of heteroge-
neous material microstructure, as we explicitly demonstrated
in this paper. In general, this framework allows one to quan-
titatively characterize and control material microstructure
via simple analytical functions. This will allow one to
parameterize the correlation functions with material process-
ing conditions (such as annealing time30), and thus establish
the quantitative processing-microstructure relation.
Structure-based analysis of the properties and performance
of the material can be carried out to establish rigorous
processing-structure-property relations. This would lead to
the development of novel material design procedure, which
we will explore in future studies. Finally, 3D material micro-
structure modeling via correlation functions is also common
in small angle x-ray scattering52 and grey-tone electron
micrograph analysis.53 It would be interesting to apply the
analytical representation scheme for correlation functions in
these situations.
ACKNOWLEDGMENTS
This work was supported by the Division of Materials
Research at National Science Foundation under Award No.
DMR-1305119 and the Office of Naval Research (ONR)
under Contract No. N00014-10-1-0350 (Dr. A. K.
Vasudevan, Program Manager). Use of the Advanced Photon
Source was supported by the U.S. Department of Energy,
Office of Science, Office of Basic Energy Sciences, under
Contract No. DE-AC02-06CH11357. The experimental as-
sistance of Jason Williams and Sudhanshu Singh at ASU,
and Xianghui Xiao and Francesco De Carlo at APS are grate-
fully acknowledged.
1S. Torquato, Random Heterogeneous Materials: Microstructure andMacroscopic Properties (Springer, New York, 2002).
2M. Sahimi, Heterogeneous Materials I: Linear Transport and OpticalProperties (Springer, New York, 2003).
3M. Sahimi, Heterogeneous Materials II: Nonlinear and BreakdownProperties and Atomistic Modeling (Springer, New York, 2003).
4A. Kak and M. Slaney, Principles of Computerized Tomographic Imaging(Springer, New York, 1988).
5D. Brandon and W. D. Kaplan, Microstructural Characterization ofMaterials (John Wiley & Sons, New York, 1999).
6G. T. Herman, Fundamentals of Computerized Tomography: ImageReconstruction From Projection, 2nd ed. (Springer, New York, 2009).
7J. Baruchel, P. Bleuet, A. Bravin, P. Coan, E. Lima, A. Madsen, W.
Ludwig, P. Pernot, and J. Susini, “Advances in synchrotron hard X-rays
based imaging,” C. R. Phys. 9, 624–641 (2008).8J. H. Kinney and M. C. Nichols, “X-ray tomographic microscopy (XTM)
using synchrotron radiation,” Annu. Rev. Mater. Sci. 22, 121–152 (1992).9E. Padilla, V. Jakkali, L. Jiang, and N. Chawla, “Quantifying the effect of
porosity on the evolution of deformation and damage in Sn-based solder
joints by x-ray microtomography and microstructure-based finite element
modeling,” Acta Mater. 60, 4017–4026 (2012).10N. Chawla, J. J. Williams, X. Deng, and C. McClimon, “Three dimen-
sional (3D) characterization and modeling of porosity in powder metal-
lurgy (P/M) steels,” Int. J. Powder Metall. 45, 19–27 (2009).11L. Babout, E. Maire, J. Y. Buffiere, and R. Fougeres, “Characterization by
X-ray computed tomography of decohesion, porosity growth and coalescence
in model metal matrix composite,” Acta Mater. 49, 2055–2063 (2001).12A. Borb�ely, F. F. Csikor, S. Zabler, P. Cloetens, and H. Biermann, “Three-
dimensional characterization of the microstructure of a metal-matrix com-
posite by holotomography,” Mater. Sci. Eng., A 367, 40–50 (2004).13P. Kenesei, H. Biermann, and A. Borb�ely, “Structure-property relationship
in particle reinforced metal-matrix composites based on holotomography,”
Scr. Mater. 53, 787–791 (2005).14J. J. Williams, Z. Flom, A. A. Amell, N. Chawla, X. Xiao, and F. De
Carlo, “Damage evolution in SiC particle reinforced Al alloy matrix com-
posites by x-ray synchrotron tomography,” Acta Mater. 58, 6194–6205
(2010).15F. A. Silva, J. J. Williams, B. R. Mueller, M. P. Hentschel, P. D. Portella,
and N. Chawla, “3D microstructure visualization of inclusions and poros-
ity in SiC particle reinforced Al matrix composites by x-ray synchrotron
tomography,” Metall. Mater. Trans. A 41, 2121–2128 (2010).16A. Weck, D. S. Wilkinson, E. Maire, and H. Toda, “Visualization by x-ray
tomography of void growth and coalescence leading to fracture in model
materials,” Acta Mater. 56, 2919–2928 (2008).17H. Toda, S. Yamamoto, M. Kobayashi, K. Uesugi, and H. Zhang, “Direct
measurement procedure for three-dimensional local crack driving force
using synchrotron X-ray microtomography,” Acta Mater. 56, 6027–6039
(2008).18M. Y. Wang, J. J. Williams, L. Jiang, F. De Carlo, T. Jing, and N. Chawla,
“Dendritic morphology of a-Mg in Mg-based alloys: Three dimensional
(3D) experimental characterization by x-ray synchrotron tomography and
phase-field simulations,” Scr. Mater. 65, 855–858 (2011).19M. Y. Wang, J. J. Williams, L. Jiang, F. De Carlo, T. Jing, and N. Chawla,
“Three dimensional (3D) microstructural characterization and quantitative
analysis of solidified microstructures in magnesium alloys by x-ray syn-
chrotron tomography,” Metallogr., Microstruct., Anal. 1, 7–13 (2012).20J. J. Williams, K. E. Yazzie, N. C. Phillips, N. Chawla, X. Xiao, F. De
Carlo, N. Iyyer, and M. Kittur, “On the correlation between fatigue stria-
tion spacing and crack growth rate: A 3D x-ray synchrotron tomography
study,” Metall. Mater. Trans. A 42, 3845–3848 (2011).21N. Chawla, V. V. Ganesh, and B. Wunsch, “Three-dimensional (3D)
microstructure visualization and finite element modeling of the mechanical
behavior of SiC particle reinforced aluminium,” Composites Scr. Mater.
51, 161–165 (2004).22N. Chawla and K. K. Chawla, “Microstructure-based modeling of defor-
mation in particle reinforced metal matrix composites,” J. Mater. Sci. 41,
913–925 (2006).23N. Chawla, R. S. Sidhu, and V. V. Ganesh, “Three dimensional (3D)
visualization and microstructure-based finite element modeling of particle
reinforced composites,” Acta Mater. 54, 1541–1548 (2006).24E. Maire, C. Bordreuil, L. Babout, and J. C. Boyer, “Damage initiation
and growth in metals. Comparison between modeling and tomography
experiments,” J. Mech. Phys. Solids 53, 2411–2434 (2005).25A. Weck, D. S. Wilkinson, and E. Maire, “Observation of void nucleation,
growth and coalescence in a model metal matrix composite using x-ray
tomography,” Mater. Sci. Eng. A 488, 435–445 (2008).
093511-8 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42
![Page 10: Modeling and characterizing anisotropic inclusion orientation in heterogeneous material via directional cluster functions and stochastic microstructure reconstruction](https://reader036.fdocuments.us/reader036/viewer/2022092616/5750a8e31a28abcf0ccc00b4/html5/thumbnails/10.jpg)
26Y. Jiao, F. H. Stillinger, and S. Torquato, “Modeling heterogeneous mate-
rials via two-point correlation functions: Basic principles,” Phys. Rev. E
76, 031110 (2007).27Y. Jiao, F. H. Stillinger, and S. Torquato, “Modeling heterogeneous mate-
rials via two-point correlation functions: II. Algorithmic details and
applications,” Phys. Rev. E 76, 031135 (2008).28Y. Jiao, F. H. Stillinger, and S. Torquato, “A superior descriptor of random
textures and its predictive capacity,” Proc. Natl. Acad. Sci. U.S.A. 106,
17634–17639 (2009).29S. S. Singh, J. J. Williams, Y. Jiao, and N. Chawla, “Modeling anisotropic
multiphase heterogeneous materials via directional correlation functions:
Simulations and experimental verification,” Metall. Mater. Trans. A 43,
4470–4474 (2012).30Y. Jiao, E. Pallia, and N. Chawla, “Modeling and predicting microstructure
evolution in lead/tin alloy via correlation functions and stochastic material
reconstruction,” Acta Mater. 61, 3370–3377 (2013).31C. L. Y. Yeong and S. Torquato, “Reconstructing random media,” Phys.
Rev. E 57, 495–506 (1998).32C. L. Y. Yeong and S. Torquato, “Reconstructing random media II. Three-
dimensional media from two-dimensional cuts,” Phys. Rev. E 58, 224–233
(1998).33Y. Liu, M. S. Greene, W. Chen, D. A. Dikin, and W. K. Liu, “Computational
microstructure characterization and reconstruction for stochastic multiscale
material design,” Comput. Aided Des. 45, 65–76 (2013).34H. Kumar, C. L. Briant, and W. A. Curtin, “Using microstructure recon-
struction to model mechanical behavior in complex microstructures,”
Mech. Mater. 38, 818–832 (2006).35S. Torquato, J. D. Beasley, and Y. C. Chiew, “Two-point cluster function
for continuum percolation,” J. Chem. Phys. 88, 6540–6547 (1988).36S. R. Niezgoda, D. T. Fullwood, and S. R. Kalidindi, “Delineation of the
space of 2-point correlations in a composite material system,” Acta Mater.
56, 5285–5292 (2008).37B. Lu and S. Torquato, “Lineal-path function for random heterogeneous
materials,” Phys. Rev. A 45, 922–929 (1992).38C. E. Zachary and S. Torquato, “Improved reconstructions of random media
using dilation and erosion processes,” Phys. Rev. E 84, 056102 (2011).39A. P. Roberts, “Statistical reconstruction of three-dimensional porous
media from two-dimensional images,” Phys. Rev. E 56, 3203–3212 (1997).40D. T. Fullwood, S. R. Niezgoda, and S. R. Kalidindi, “Microstructure
reconstructions from 2-point statistics using phase-recovery algorithms,”
Acta Mater. 56, 942–948 (2008).
41P. Tahmasebi and M. Sahimi, “Cross-correlation function for accurate
reconstruction of heterogeneous media,” Phys. Rev. Lett. 110, 078002
(2013).42S. Kirkpatrick, C. D. Gelatt, and M. P. Vecchi, “Optimization by simulated
annealing,” Science 220, 671–680 (1983).43D. M. Saylor, J. Fridy, B. S. El-Dasher, K. Y. Jung, and A. D. Rollet.
“Statistically representative three-dimensional microstructures based on
orthogonal observation sections,” Metall. Mater. Trans. A 35A,
1969–1979 (2004).44A. Brahme, M. H. Alvi, D. Saylor, J. Fridy, and A. D. Rollet, “3D recon-
struction of microstructure in a commercial purity aluminum,” Scr. Mater.
55, 75–80 (2006).45M. Groeber, S. Ghosh, M. D. Uchic, and D. M. Dimiduk, “A framework
for automated analysis and simulation of 3D polycrystalline microstruc-
tures. Part I: Statistical characterization,” Acta Mater. 56, 1257–1273
(2008).46M. Groeber, S. Ghosh, M. D. Uchic, and D. M. Dimiduk, “A framework
for automated analysis and simulation of 3D polycrystalline microstruc-
tures. Part II: Synthetic structure generation,” Acta Mater. 56, 1274–1287
(2008).47J. J. Williams, K. E. Yazzie, E. Padilla, N. Chawla, X. Xiao, and F. De
Carlo, “Understanding fatigue crack growth in aluminum alloys by in situ
x-ray synchrotron tomography,” Int. J. Fatigue 57, 79–85 (2013).48P. Debye, H. R. Anderson, and H. Brumberger, “Scattering by an inhomo-
geneous solid. II. The correlation function and its applications,” J. Appl.
Phys. 28, 679–683 (1957).49Y. Jiao, F. H. Stillinger, and S. Torquato, “Geometrical ambiguity of pair
statistics. II. Heterogeneous media,” Phys. Rev. E 82, 011106 (2010).50C. J. Gommes, Y. Jiao, and S. Torquato, “Density of states for a specified
correlation function and the energy landscape,” Phys. Rev. Lett. 108,
080601 (2012).51S. Torquato and F. H. Stillinger, “Jammed hard-particle packings:
From Kepler to Bernal and beyond,” Rev. Mod. Phys. 82, 2633–2672
(2010).52C. J. Gommes, “Three-dimensional reconstruction of liquid phases in dis-
ordered mesopores using in situ small-angle scattering,” J. Appl. Cryst. 46,
493–504 (2013).53C. J. Gommes, H. Friedrich, P. E. de Jongh, and K. P. de Jong, “2-Point
correlation function of nanostructured materials via the grey-tone correla-
tion function of electron tomograms: A three-dimensional structure analy-
sis of ordered mesoporous media,” Acta Mater. 58, 770–780 (2010).
093511-9 Y. Jiao and N. Chawla J. Appl. Phys. 115, 093511 (2014)
[This article is copyrighted as indicated in the article. Reuse of AIP content is subject to the terms at: http://scitation.aip.org/termsconditions. Downloaded to ] IP:
141.217.58.222 On: Wed, 26 Nov 2014 19:05:42